TY - JOUR
T1 - Improving field boundary delineation in ResUNets via adversarial deep learning
AU - Jong, Maxwell
AU - Guan, Kaiyu
AU - Wang, Sibo
AU - Huang, Yizhi
AU - Peng, Bin
N1 - This work was supported in part by a grant from NSF CAREER Award (1847334), USDA NIFA Foundational Program, and the Advanced Research Projects Agency - Energy (ARPA-E) Phase II grant funded by the US Department of Energy (DOE) (DOE DE-AR0001382).
PY - 2022/8
Y1 - 2022/8
N2 - Field boundary data is often required to access digital agricultural services and tools that assist with field-level assessment and monitoring. In addition, policy-makers and researchers need field boundaries to accurately assess food security and impacts on climate change. Thus, scalable and efficient automatic field boundary detection algorithms on satellite images have direct, important implications for many stakeholders. Deep learning is one approach that has been successfully applied in recent years to field boundary detection. Qualitatively however, these boundaries are often broken or malformed, necessitating a dependence on fine-tuned post-processing methods with arbitrary thresholds obtained through trial-and-error. Prior work has explored various architectures for predicting field boundaries, but little has been done beyond traditional supervised learning regimes. Thus, in this work, we propose a new approach to improving field boundary prediction by using an adversarial training framework. In particular, we investigated the effects of training a ResUNet model (a standard fully convolutional network architecture) as a generator in a traditional generative adversarial network (GAN) setup, on 30 meter resolution satellite imagery from 2017 over the state of Illinois. We then explored whether or not our methods can be transferred to label-scarce regions in Brazil. Overall, our results showed that adversarial training substantially improved boundary quality and performance, but had a lesser effect when transferred to unseen, low-data agricultural landscapes. Based on these findings, we conclude that adversarial training is a promising way to improve boundary quality during prediction time, and we suggest several ideas for future improvements that may make adversarial training more viable in transfer learning.
AB - Field boundary data is often required to access digital agricultural services and tools that assist with field-level assessment and monitoring. In addition, policy-makers and researchers need field boundaries to accurately assess food security and impacts on climate change. Thus, scalable and efficient automatic field boundary detection algorithms on satellite images have direct, important implications for many stakeholders. Deep learning is one approach that has been successfully applied in recent years to field boundary detection. Qualitatively however, these boundaries are often broken or malformed, necessitating a dependence on fine-tuned post-processing methods with arbitrary thresholds obtained through trial-and-error. Prior work has explored various architectures for predicting field boundaries, but little has been done beyond traditional supervised learning regimes. Thus, in this work, we propose a new approach to improving field boundary prediction by using an adversarial training framework. In particular, we investigated the effects of training a ResUNet model (a standard fully convolutional network architecture) as a generator in a traditional generative adversarial network (GAN) setup, on 30 meter resolution satellite imagery from 2017 over the state of Illinois. We then explored whether or not our methods can be transferred to label-scarce regions in Brazil. Overall, our results showed that adversarial training substantially improved boundary quality and performance, but had a lesser effect when transferred to unseen, low-data agricultural landscapes. Based on these findings, we conclude that adversarial training is a promising way to improve boundary quality during prediction time, and we suggest several ideas for future improvements that may make adversarial training more viable in transfer learning.
KW - Adversarial learning
KW - Boundary quality
KW - Field boundaries
KW - Semantic segmentation
KW - Transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85133342740&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85133342740&partnerID=8YFLogxK
U2 - 10.1016/j.jag.2022.102877
DO - 10.1016/j.jag.2022.102877
M3 - Article
AN - SCOPUS:85133342740
SN - 1569-8432
VL - 112
JO - International Journal of Applied Earth Observation and Geoinformation
JF - International Journal of Applied Earth Observation and Geoinformation
M1 - 102877
ER -